BigScience and the BLOOM Model Quiz

Answers and explanations appear as you take the quiz.

1. What was the primary distinguishing goal of the BLOOM model compared to similar-sized models of its time?
   • Being multilingual and openly released
   • Being the fastest at inference
   • Being the smallest possible model
   • Being trained only on programming code
2. Approximately how many parameters does BLOOM have?
   • 7 billion
   • 176 billion
   • 70 billion
   • 540 billion
3. How many natural languages was BLOOM trained to handle?
   • 1
   • 100
   • 46
   • 600
4. Which organization coordinated the BigScience collaboration?
   • DeepMind
   • Stability AI
   • Anthropic
   • Hugging Face
5. Where was BLOOM trained?
   • On the publicly funded Jean Zay supercomputer in France
   • On private cloud GPUs rented from a startup
   • On a university laptop cluster
   • On Google's TPU pods
6. How does BLOOM’s ALiBi mechanism encode relative position?
   • It removes attention and uses a convolutional decoder
   • It adds distance-dependent biases to attention scores
   • It inserts learned position vectors into every token embedding
   • It stores position as a separate image channel
7. How many programming languages were represented in BLOOM’s ROOTS corpus?
   • 3 programming languages
   • 46 programming languages
   • 13 programming languages
   • 130 programming languages
8. What model architecture does the BLOOM paper describe?
   • A convolutional image classifier
   • A recurrent speech recognizer
   • A diffusion model for audio
   • A decoder-only Transformer